*******************************************************************************************************************************************************
	*SETTING GLOBALS FOR THE ANALYSIS
*******************************************************************************************************************************************************
global data_raw "$cd/data/1_raw"
global data_inter "$cd/data/2_intermediate"
global data_final "$cd/data/3_final"

global outtex "$cd/output/regressions/"
global outfig "$cd/output/figures/"
global dopath "$cd/do files"
global log "$cd/log"

	
*******************************************************************************************************************************************************
	*TABLE SHELL SETTINGS
*******************************************************************************************************************************************************

	global prehead1 "\begin{tabular}{l*{"
	global prehead2 "}{c}} \hline\hline"
	global weight " "


*******************************************************************************************************************************************************
	*FOOTNOTES
*******************************************************************************************************************************************************		
		
	
	global footnote_s1 "{\bf Specification:} This table estimates Specification \ref{eq:1-1-1} in the paper. In Panel A, Average Rank indicates the {\it average} ranking the entrepreneur was given by her peers for the question in the column heading. In Panel B, Average Rank indicates the percentile of Average Rank Level. The Average Rank is computed excluding a person's own self rank. In columns (1), (2), and (3), the number of observations is greater than the number of households because we regress the outcome on both the zero sum (relative) and the non-zero sum (quintile) rank in a stacked regression and control for the ranking question. All respondents were asked to provide the quintile and relative rank for a randomly selected two of these three questions. A subset of respondents were also asked to provide the relative rank for the third question. A subset of respondents were also randomly selected to provide the relative rank for the questions in columns (4)- (6).  Robust standard errors clustered at the group level in parentheses. All regressions include randomization strata, survey month, survey round, and surveyor fixed effects. The analogue of this table that includes the self rank can be found in Table \ref{avggrespondentsknowwithself}. " 

	global footnote_o1 "\\  {\bf Outcome variables:}  In Panel A, the outcome variable is the level of the outcome labeled in the column header, as reported by the rankee at baseline. In Panel B, the outcome variable is the percentile of the outcome in Panel B. The number of observations varies across questions because each respondent answered only a subset of the questions as explained in Section \ref{subsec:Elicitation-Exercise}. For a description of the data that produced the outcome variables, see the Appendix \ref{implementationappendix}."
	
	global footnote_s2 "{\bf Specification:} This table estimates Specification \ref{eq:1} in the paper. Rank indicates the {\it average} ranking the entrepreneur was given by her peers for the marginal returns to grant quintile ranking (non-zero sum) question. It excludes the self rank before producing the average ranking.  See Figure \ref{fig:mr_distribution} for a distribution of average rank.  Top (Middle) Tercile Rank is a dummy for whether the entrepreneur is in the top (middle) tercile of the average marginal return rank distribution. Winner indicates that the household is a grant recipient after baseline (after round 1 of data collection). The unit of observation is the household. Robust standard errors clustered at the group level in parentheses. All regressions include household, survey month, survey round, and surveyor fixed effects. The even columns also include all of the baseline controls in Table \ref{balance} interacted with Winner. All regressions are weighed by the inverse propensity score described in Section \ref{subsec:Average-Returns}." 

	global footnote_o2 "\\ {\bf Outcome variables:} In columns (1)-(2) and (5)-(6) we show the {\it trimmed} distributions of income and profits, respectively, as described in Section \ref{subsec:Average-Returns}. In columns (3)-(4) and (7)-(8), we show the natural log of the (outcome+1) of the {\it untrimmed} distribution (which is why the number of observations is greater than in the preceding column).  For a description of the data that produced the outcome variables, see the Appendix \ref{implementationappendix}."

	global footnote_d2 "Data in this table come from rounds 1-4 of data collection."
	
	global footnote_s4 "{\bf Specification:} This table estimates Specification 8 in the paper. Top (middle) Tercile Controls is a dummy for whether the entrepreneur is in the top (middle) tercile of predicted marginal return to capital based on observables. Top (middle) Tercile Controls+Rank is a dummy for whether the entrepreneur is in the top (middle) tercile of predicted marginal return to capital based on observables plus the average community ranking (excluding the entrepreneur's ranking of herself).  Both predictive models were constructed using the process described in Section 4.4. Winner indicates that the household is a grant recipient after baseline (after round 1 of data collection). The unit of observation in the household. Robust standard  errors clustered at the group level in parentheses. All regressions include household, survey month, survey round, and surveyor fixed effects. The even columns also include all of the baseline controls in Table \ref{balance} interacted with Winner. All regressions are weighed by the inverse propensity score described in Section \ref{subsec:Average-Returns}."
	
	global footnote_s3 "{\bf Specification:} This table estimates Specification \ref{eq:1} in the paper. Rank indicates the {\it average} ranking the entrepreneur was given by her peers for the marginal returns to grant quintile ranking (non-zero sum) question. It excludes the self rank before producing the average ranking.  See Figure \ref{fig:mr_distribution} for a distribution of average rank.  Top (Middle) Tercile Rank is a dummy for whether the entrepreneur is in the top (middle) tercile of the average marginal return rank distribution. Winner indicates that the household is a grant recipient after baseline (after round 1 of data collection). The unit of observation is the household. Robust standard errors clustered at the group level in parentheses. All regressions include household, survey month, survey round, and surveyor fixed effects.  All regressions are weighed by the inverse propensity score described in Section \ref{subsec:Average-Returns}." 
	
	global footnote_o3 "\\ {\bf Outcome variables:} The number of observations in columns 1-4 varies due to missing data across the rounds. Variables reported in columns 5-10 were only collected at baseline and in round 4. For a description of the data that produced the outcome variables, see the Appendix \ref{implementationappendix}. "
	
	
	global footnote_sa1  " {\bf Specification:} This table estimates Specification \ref{eq:0} in the paper. The even columns show the coefficient \$\tau_{1}\$ from that regression model and the Treatment is specified in the column heading. The odd columns show the mean for persons in the control group of the treatment described in the column heading. For example, the top row of column 1 shows the probability of being a male for persons assigned to the No Stakes group. The value in column 2 is the difference in the probability of being male for persons assigned to the High Stakes group.  Standard errors are clustered at group level. The model includes randomization strata fixed effects. "  

	global footnote_da1 "Data in this table come from round 1 of data collection."
	
	global footnote_oa1 "\\ {\bf Outcome variables:} The characteristics in Panels A and B are of the entrepreneur and her main businesses that was ranked in the elicitation exercise. The characteristics in Panel C are for the entrepreneur's household. In Panel D, we show business characteristics summed across all household businesses. If the household only has one business, then these are the summary statistics for that business. For details of how the outcome variables are constructed, see the Appendix \ref{implementationappendix}"
		

		global footnote_sa2  "{\bf Specification:} This table estimates Specification \ref{eq:1-1-1} in the paper. In Panel A, Average Rank indicates the {\it average} ranking the entrepreneur was given by her peers for the question in the column heading. In Panel B, Average Rank indicates the percentile of Average Rank Level. The Average Rank is computed {\it including} a person's own self rank. In columns (1), (2), and (3), the number of observations is greater than the number of households because we regress the outcome on both the zero sum (relative) and the non-zero sum (quintile) rank in a stacked regression and control for the ranking question. All respondents were asked to provide the quintile and relative rank for a randomly selected two of these three questions. A subset of respondents were also asked to provide the relative rank for the third question. A subset of respondents were also randomly selected to provide the relative rank for the questions in columnd (4)- (6).  Robust standard errors clustered at the group level in parentheses. All regressions include randomization strata, survey month, survey round, and surveyor fixed effects."
		global footnote_oa2 "\\ {\bf Outcome variables:}  In Panel A, the outcome variable is the level of the outcome labeled in the column header, as reported by the rankee at baseline. In Panel B, the outcome variable is the percentile of the outcome in Panel B. The number of observations varies across questions because each respondent answered only a subset of the questions as explained in Section \ref{subsec:Elicitation-Exercise}. For a description of the data that produced the outcome variables, see the Appendix \ref{implementationappendix}." 
	
	
	global footnote_sa3 "{\bf Specification:} This table estimates Specification \ref{eq:1-2} in the paper. Winner indicates that the household is a grant recipient after baseline (after round 1 of data collection). The unit of observation is the household. Robust standard errors clustered at the group level in parentheses. All regressions include household, survey month, survey round, and surveyor fixed effects. All regressions are weighed by the inverse propensity score described in Section \ref{subsec:Average-Returns}." 
	
	global footnote_oa3 "\\ {\bf Outcome variables:} In columns (1) and (3) we show the {\it trimmed} distributions of income and profits, respectively, as described in Section \ref{subsec:Average-Returns}. In columns (2) and (4), we show the natural log of the (outcome+1) of the {\it untrimmed} distribution (which is why the number of observations is greater than in the preceding column).  For a description of the data that produced the outcome variables, see the Appendix \ref{implementationappendix}."
	
	
	global footnote_sa4  "{\bf Specification:} This table estimates Specification \ref{eq:0} in the paper in order to conduct a balance test of grant randomization by average marginal returns rank tercile. The even columns show the coefficient \$\tau_{1}\$ from that regression model and the treatment Treatment, in this case grant winner. The odd columns show the mean for persons in the control group. To produce columns 1 and 2, we limit the sample to persons who were ranked in the top tercile of the average marginal returns rank distribution. As an example, the top row of column 1 shows the probability of being a male for persons who were ranked in the top tercile of the average marginal returns rank distribution and did not win the grant. The value in column 2 is the difference in the probability of being male for persons who were ranked in the top tercile of the average marginal returns rank distribution and did win the grant.  Standard errors are clustered at group level. The model includes randomization strata fixed effects. " 

	
	global footnote_sa5 "{\bf Specification:} This table estimates Specification \ref{eq:1-1} in the paper, without interaction term Winner*Rank.  Winner indicates that the household is a grant recipient. We limit the regression to post-grant distribution data collection rounds (round 2-4) and we control for the baseline value of the outcome (from round 1). The unit of observation is the household. Robust standard errors clustered at the group level in parentheses. All regressions include randomization strata, survey month, survey round, and surveyor fixed effects. All regressions are weighed by the inverse propensity score described in Section \ref{subsec:Average-Returns}." 	
	
	global footnote_oa5 "\\ {\bf Outcome variables:} In columns (1) and (3) we show the {\it trimmed} distributions of income and profits, respectively, as described in Section \ref{subsec:Average-Returns}. In columns (2) and (4), we show the natural log of the (outcome+1) of the {\it untrimmed} distribution (which is why the number of observations is greater than in the preceding column).  For a description of the data that produced the outcome variables, see the Appendix \ref{implementationappendix}." 	
	
	
	global footnote_sa6 "{\bf Specification:} This table estimates Specification \ref{eq:1-1} in the paper.  Winner indicates that the household is a grant recipient. We limit the regression to post-grant distribution data collection rounds (round 2-4) and we control for the baseline value of the outcome (from round 1). The unit of observation is the household. Robust standard errors clustered at the group level in parentheses. All regressions include randomization strata, survey month, survey round, and surveyor fixed effects. All regressions are weighed by the inverse propensity score described in Section \ref{subsec:Average-Returns}." 		
	
	
	global footnote_sa7 "{\bf Specification:} This table estimates Specification \ref{eq:1} in the paper. Rank indicates the {\it average} ranking the entrepreneur was given by her peers for the marginal returns to grant quintile ranking (non-zero sum) question. Analysis is constrained to only groups of 5 members and always excludes the self rank before producing the average ranking.  So there are 4 peer reports per group. In column (1), the rank is averaged over 4 peer reports. There is one observation per rankee since there are 4 total reports. In column (2), all combinations of 3 peer reports are averaged; so there are 4 averaged ranks reports per rankee. In column (3), all pairs of peer reports are averaged, so there are 6 ranks reports per rankee.  In column (4), all reports are individually analyzed; so there are 4 reports peer rankee. At the bottom of the table, we show the p-value from an f-test of the coefficient for Winner*Rank (averaged over 4 reports) and Winner*Rank in each subsequent column. Winner indicates that the household is a grant recipient after baseline (after round 1 of data collection). The unit of observation is the household. Robust standard errors clustered at the group level in parentheses. All regressions include household, survey month, survey round, and surveyor fixed effects.  All regressions are weighed by the inverse propensity score described in Section \ref{subsec:Average-Returns}." 	
	
	global footnote_oa7 "\\ {\bf Outcome variables:} In columns (1)-(4) we show the {\it trimmed} distributions of profits. For a description of the data that produced the outcome variables, see the Appendix \ref{implementationappendix}."
	
	
	global footnote_sa8 "{\bf Specification:} This table estimates Specification \ref{eq:1} in the paper. Rank indicates the {\it average} ranking the entrepreneur was given by her peers for the marginal returns to grant quintile ranking (non-zero sum) question. Unlike Table \ref{rankmreffectfe}, average rank includes the self rank before producing the average ranking. Top (Middle) Tercile Rank is a dummy for whether the entrepreneur is in the top (middle) tercile of the average marginal return rank distribution. Winner indicates that the household is a grant recipient after baseline (after round 1 of data collection). The unit of observation is the household. Robust standard errors clustered at the group level in parentheses. All regressions include household, survey month, survey round, and surveyor fixed effects. The even columns also include all of the baseline controls in Table \ref{balance} interacted with Winner. All regressions are weighed by the inverse propensity score described in Section \ref{subsec:Average-Returns}."	
	
	
	global footnote_sa9 "{\bf Specification:} This table estimates Specification \ref{eq:1} in the paper. Unlike in Table \ref{rankmreffectfe} Rank indicates the average ranking the entrepreneur was given by her peers for the marginal returns to grant {\it relative} ranking (zero sum) question. It excludes the self rank before producing the average ranking. Top (Middle) Tercile Rank is a dummy for whether the entrepreneur is in the top (middle) tercile of the average marginal return rank distribution. Winner indicates that the household is a grant recipient after baseline (after round 1 of data collection). The unit of observation is the household. Robust standard errors clustered at the group level in parentheses. All regressions include household, survey month, survey round, and surveyor fixed effects. The even columns also include all of the baseline controls in Table \ref{balance} interacted with Winner. All regressions are weighed by the inverse propensity score described in Section \ref{subsec:Average-Returns}." 	
	
	global footnote_sa10 "{\bf Specification:} This table estimates Specification \ref{eq:1} in the paper. Unlike in Table \ref{rankmreffectfe} Median Rank indicates the {\it median} (rather than average) ranking the entrepreneur was given by her peers for the marginal returns to grant quintile ranking (non-zero sum) question. It excludes the self rank before producing the average ranking. Top (Middle) Tercile Rank is a dummy for whether the entrepreneur is in the top (middle) tercile of the median marginal return rank distribution. Winner indicates that the household is a grant recipient after baseline (after round 1 of data collection). The unit of observation is the household. Robust standard errors clustered at the group level in parentheses. All regressions include household, survey month, survey round, and surveyor fixed effects. All regressions are weighed by the inverse propensity score described in Section \ref{subsec:Average-Returns}."
	
	global footnote_sa11 "{\bf Specification:} Rank indicates the {\it average} ranking the entrepreneur was given by her peers for the marginal returns to grant quintile ranking (non-zero sum) question. Top (Middle) Tercile Rank is a dummy for whether the entrepreneur is in the top (middle) tercile of the average marginal return rank distribution. Winner indicates that the household is a grant recipient after baseline (after round 1 of data collection). It excludes the self rank before producing the average ranking. Std Rank is the standard deviation of Rank. Winner indicates that the household is a grant recipient after baseline (after round 1 of data collection). In Panel A, we interact Std Rank with the linear rank. In Panel B, the same measure is interacted with top and middle tercile. The unit of observation is the household. Robust standard errors clustered at the group level in parentheses. All regressions include household, survey month, survey round, and surveyor fixed effects. All regressions are weighed by the inverse propensity score described in Section \ref{subsec:Average-Returns}."
	

	global footnote_sa12 "{\bf Specification:} This table estimates Specification \ref{eq:1} in the paper. Rank indicates the {\it average} ranking the entrepreneur was given by her peers for the marginal returns to grant quintile ranking (non-zero sum) question. It excludes the self rank before producing the average ranking. Top (Middle) Tercile Rank is a dummy for whether the entrepreneur is in the top (middle) tercile of the average marginal return rank distribution. Winner indicates that the household is a grant recipient after baseline (after round 1 of data collection). Unlike in Table \ref{rankmreffectfe}, the unit of observation is the entrepreneur that was ranked during the ranking exercise (rather than the household).  We aggregate across all of the businesses owned by that entrepreneur (as opposed to in previous tables where we aggregate across all businesses owned by the household). Robust standard errors clustered at the group level in parentheses. All regressions include household, survey month, survey round, and surveyor fixed effects. The even columns also include all of the baseline controls in Table \ref{balance} interacted with Winner. All regressions are weighed by the inverse propensity score described in Section \ref{subsec:Average-Returns}." 	
	
	
	global footnote_oa13 "\\ {\bf Outcome variables:} In columns (1)-(2) and (3)-(4) we show the trimmed distributions of income and profits  adjusted for the owner's own labor costs.  To compute the value of the owner's own labor, we first create an estimated daily wage value for each entrepreneur by the entrepreneur's education and gender. This daily wage is multiple by each entrepreneur's days worked over the previous 30 days. This value is then subtracted from the profits of the entrepreneur's business (columns 3 and 4) or from total household income (columns 1 and 2).  For a description of the data that produced the outcome variables, see the Appendix \ref{implementationappendix}."		
	
	
	
	global footnote_da14 "Unlike in Table \ref{rankmreffectfe}, data in this table come from rounds 1-5 of data collection."		
	
	
	global footnote_sa15 "{\bf Specification:} This table estimates Specification \ref{eq:1} in the paper. Rank indicates the {\it average} ranking the entrepreneur was given by her peers for the marginal returns to grant quintile ranking (non-zero sum) question. It excludes the self rank before producing the average ranking.  Unlike in Table \ref{rankmreffectfe}, analysis in this table is limited to only groups of 5 entrepreneurs. For a distribution of the number of entrepreneurs per group, see Figure \ref{fig:groupsize}. Top (Middle) Tercile Rank is a dummy for whether the entrepreneur is in the top (middle) tercile of the average marginal return rank distribution. Winner indicates that the household is a grant recipient after baseline (after round 1 of data collection). The unit of observation is the household. Robust standard errors clustered at the group level in parentheses. All regressions include household, survey month, survey round, and surveyor fixed effects. The even columns also include all of the baseline controls in Table \ref{balance} interacted with Winner. All regressions are weighed by the inverse propensity score described in Section \ref{subsec:Average-Returns}." 
	
	
	global footnote_sa16 "{\bf Specification:} The sample is limited to households that received the grant. In Panel A, we regress amount of the Rs. 6000 grant that the household spent in a particular category on the Rank. Rank indicates the {\it average} ranking the entrepreneur was given by her peers for the marginal returns to grant quintile ranking (non-zero sum) question. It excludes the self rank before producing the average ranking.  Top (Middle) Tercile Rank is a dummy for whether the entrepreneur is in the top (middle) tercile of the quintile (non-zero sum) average marginal return rank distribution.  Robust standard errors clustered at the group level in parentheses. All regressions include randomization strata, survey month, survey round, and surveyor fixed effects.   "
	
	global footnote_da16 "Households were asked how the invested the grant every survey round. Households who say the grant was saved (column 10) are households that by the last survey round still had not spent the grant amount."
	
	global footnote_oa16 "\\ {\bf Outcome variables:} Respondents were asked to report how they spent the grant they received. In column (1), we report the results from the question Did you add any of your own money to the grant amount to make a purchase? Column (2) is the sum of columns (3)-(6). Column (7) is the sum of columns (8)-(9). For a description of the data that produced the outcome variables, see the Appendix \ref{implementationappendix}"
	
	global footnote_sa17 "{\bf Specification:} This table estimates Specification \ref{eq:1} in the paper. Rank indicates the {\it average} ranking the entrepreneur was given by her peers for the marginal returns to grant quintile ranking (non-zero sum) question. Top (Middle) Tercile Rank is a dummy for whether the entrepreneur is in the top (middle) tercile of the average marginal return rank distribution. Winner indicates that the household is a grant recipient after baseline (after round 1 of data collection). The unit of observation is the household. Robust standard errors clustered at the group level in parentheses. All regressions include household, survey month, survey round, and surveyor fixed effects. The even columns also include baseline controls interacted with Winner. They also include Winner interacted with each of the 17 psychometric questions elicited at baseline (see Appendix \ref{psych}). Specifically, Winner is interacted with a dummy for each of the 17 questions if the response to the question was Strongly Agree.   All regressions are weighed by the inverse propensity score described in Section \ref{subsec:Average-Returns}."
	
	
	
	
****************************************************************************************************************
****************************************************************************************************************
****************************************************************************************************************
	
	global footnote_s5 "{\bf Specification:} This table estimates Specification \ref{eq:1-1-1-1} in the paper. The regressions include Stakes, but the coefficient is not reported in the table. In columns (1)-(3), Rank is the percentile corresponding to the rank that person {\it i} in the group assigned to entrepreneur {\it j} in the group. So the unit of observation in these 3 columns is the ranker-rankee pair. Rank excludes the self rank. In columns (4)-(6), Average Rank indicates the percentile of  the {\it average} ranking the entrepreneur was given by her peers for a particular question. So the unit of observation is the rankee. Average Rank excludes the self rank. Robust standard errors clustered at the group level in parentheses. All regressions include ranking question, randomization strata, survey month, and surveyor fixed effects. The analogue of this table that includes the self rank can be found in Table \ref{respondentsliewithself}." 
	
	global footnote_o5 "\\ {\bf Outcome variables:}  In columns (1) and (4), we pool across questions (1)-(3) in Panel A of Table 1 (in order to be comparable across questions, the outcome variable is percentilized). In columns (2) and (5), we limit the analysis the quintile (non-zero sum) questions. In columns (3) and (6), we limit the analysis to the relative (zero-sum) questions. So column (1) pools columns (2) and (3) together. Column (4) pools columns (5) and (6) together. The number of observations varies between columns (2) and (3) because because each respondent answered only a subset of the questions as explained in Section \ref{subsec:Elicitation-Exercise}. For a description of the data that produced the outcome variables, see the Appendix \ref{implementationappendix}." 
	
	
	global footnote_sa20 "{\bf Specification:} This table estimates Specification \ref{eq:1-1-1-1} in the paper.  The regressions include Stakes, but the coefficient is not reported in the table. In columns (1)-(3), Rank is the percentile corresponding to the rank that person {\it i} in the group assigned to entrepreneur {\it j} in the group. So the unit of observation in these 3 columns is the ranker-rankee pair. Rank {\it includes} the self rank. In columns (4)-(6), Average Rank indicates the percentile of  the {\it average} ranking the entrepreneur was given by her peers for a particular question. So the unit of observation is the rankee. Average Rank excludes the self rank. Robust standard errors clustered at the group level in parentheses. All regressions include ranking question, randomization strata, survey month, and surveyor fixed effects. " 
	
	global footnote_s6 "{\bf Specification:} This table estimates Specification \ref{eq:1-1-1-2} in the paper. The regressions include Incentives, Public, and Incentives*Public, but the coefficients are not reported in the table. Average Rank in columns (1) and (2) is the percentile of the rank that an entrepreneur assigns to herself on a particular question. In columns (3) and (4), Average Rank is the percentile of the {\it average} ranking the entrepreneur was given by her peers for a particular question (excluding the rank she assigned to herself). The unit of observation is the rankee by question. In columns (1) and (3), we limit the analysis to the {\it No Stakes} treatment group. In columns (2) and (4), we limit the analysis to the {\it Stakes} group. All regressions include ranking question, randomization strata, survey month, and surveyor fixed effects. " 
	
	global footnote_o6 "\\ {\bf Outcome variables:}  We pool across questions (1)-(3) in Panel A of Table 1 (in order to be comparable across questions, the outcome variable is percentilized) and that is the outcome across all columns of the table. For a description of the data that produced the outcome variables, see the Appendix \ref{implementationappendix}." 
	
	
	
	global footnote_sa21 "{\bf Specification:} We interact the treatment status (incentives, public)  with the member indicated in the row Characteristic at the top of the table. The regressions include Incentives, Public, and Incentives*Public, but the coefficients are not reported in the table. In columns (1)-(3), the interaction is with a dummy for whether the ranker is a family member of the rankee.  In columns (4)-(6), the interaction is with a dummy for whether the ranker is close peer of the rankee (as reported by other members of the group).  In columns (2) and (5), we limit the analysis to the {\it Stakes} treatment group. In columns (3) and (6), we limit the analysis to the {\it No Stakes} group.  In columns (1) and (3), we pool across the two. The unit of observation is the ranker-rankee pair. Robust standard errors clustered at the group level in parentheses. All regressions include ranking question, randomization strata, survey month, and surveyor fixed effects. " 	
	
	global footnote_oa21 "\\ {\bf Outcome variables:}  The outcome variable is the rank that that person {\it i} in the group assigned to entrepreneur {\it j} in the group. For a description of the data that produced the outcome variables, see the Appendix \ref{implementationappendix}."
	

	global footnote_s22 "{\bf Specification:} This table estimates Specification \ref{eq:1-1-1-2} in the paper, but excludes the interaction Average Rank*Public*Incentive as well as Public*Incentive. The regressions include Incentives and Public, but the coefficients are not reported in the table. Average Rank in columns (1) and (2) is the percentile of the rank that an entrepreneur assigns to herself on a particular question. In columns (3) and (4), Average Rank is the percentile of the {\it average} ranking the entrepreneur was given by her peers for a particular question (excluding the rank she assigned to herself). The unit of observation is the rankee by question. In columns (1) and (3), we limit the analysis to the {\it No Stakes} treatment group. In columns (2) and (4), we limit the analysis to the {\it Stakes} group. All regressions include ranking question, randomization strata, survey month, and surveyor fixed effects. " 
	
		global footnote_sa23 "{\bf Specification:} This table estimates Specification \ref{eq:characetristic_knowledge} in the paper. Rank is the percentile of the ranking the entrepreneur was given by her peers for a particular question (excluding the rank she assigned to herself). The unit of observation is the ranker by rankee by question. In column (1) the sample is restricted to male rankees, and Characteristic is a dummy for whether the ranker is male. In column (2) the sample is restricted to female rankees, and Characteristic is a dummy for whether the ranker is female. In column (3) the sample is restricted to tailor rankees, and Characteristic is a dummy for whether the ranker is a tailor. In column (4) the sample is restricted to vegetable vendor rankees, and Characteristic is a dummy for whether the ranker is a vegetable vendor. In column (5) the sample is restricted to kirana shop rankees, and Characteristic is a dummy for whether the ranker is a kirana shop owner. All regressions include ranking question, randomization strata, survey month, and surveyor fixed effects." 
	
	global footnote_oa23 "\\ {\bf Outcome variables:} We pool across questions in columns (2), (5), and (6) in Panel A of Table 1 (in order to be comparable across questions, the outcome variable is percentilized) - and that is the outcome across all columns of the table.  We limit the responses to these questions as these are the ones measured at the level of the entrepreneur, rather than the household.  For a description of the data that produced the outcome variables, see the Appendix \ref{implementationappendix}."
	
	global footnote_sa24 "{\bf Specification:} This table estimates Specification \ref{eq:cross_report} in the paper. The regressor (Rank) is the percentile corresponding to the rank that person {\it i} in the group assigned to entrepreneur {\it j} in the group (it exludes entrepreneurs' rank about themselves). The regressions include Most Informed, but the coefficients are not reported in the table. So the unit of observation is the ranker-rankee pair.  Rank is interacted with Most Informed, which is a dummy variable that indicates whether at least 3 group members agree that the ranker has the most information to answer a particular ranking question. Robust standard errors clustered at the group level in parentheses. All regressions include ranking question, randomization strata, survey month, and surveyor fixed effects.  "

	global footnote_oa24 "\\ {\bf Outcome variables:}   We pool across all questions in Panel A of Table 1 (in order to be comparable across questions, the outcome variable is percentilized) and that is the outcome across all columns of the table. For a description of the data that produced the outcome variables, see the Appendix \ref{implementationappendix}."	

*****************************************************************************************************************************************************************************
* FOOTNOTES ADDED TO REPLICATE TABLES IN "Targeting High Ability Entrepreneurs: Replication and Heterogeneity by Gender of Hussam, Rigol and Roth (2022)"
*****************************************************************************************************************************************************************************
	
	global footnote_t4_winsorizing "{\bf Notes:} This table replicates Table 4 in Hussam et al. (2022b) but replacing \textit{trimmed} income and profits in Columns 1-2 and 5-6 with \textit{winsorized} incomes and profits. Winsorized outcomes are obtained by replacing observations above the 99.5th percentile by the value of outcome at the 99.5th percentile."
	
	global footnote_t2_winsorizing "{\bf Notes:} This table replicates Table 2 in Hussam et al. (2022b) but adding  \textit{winsorized} income and profits in Columns 3-4 and 9-10, respectively. Winsorized outcomes are obtained by replacing observations above the 99.5th percentile by the value of outcome at the 99.5th percentile."
	
	global footnote_t4_IHS "{\bf Notes:} This table replicates Table 4 in Hussam et al. (2022b) but using the inverse hyperbolic sine (IHS) transformation instead of a logarithmic transformation for dependent variables in Columns 3-4 and 7-8."
	
	global footnote_t2_IHS "{\bf Notes:} This table replicates Table 2 in Hussam et al. (2022b) but adding the IHS transformation for income and profits in Columns 5-6 and 11-12, respectively."
	
	global footnote_t4_grouping "{\bf Notes:} This table replicates Table 4 in Hussam et al. (2022b) but dividing the average rank distribution of entrepreneurs into quartiles instead of terciles."
	
	global footnote_t2_grouping "{\bf Notes:} This table replicates Table 2 in Hussam et al. (2022b) but dividing the average rank distribution of entrepreneurs into quartiles instead of terciles."
	
	global footnote_t4_removingFEs "{\bf Notes:} This table replicates Table 4 in Hussam et al. (2022b) but dropping surveyor and survey month fixed effects."
	
	global footnote_t2_removingFEs "{\bf Notes:} This table replicates Table 2 in Hussam et al. (2022b) but dropping surveyor and survey month fixed effects."

	global footnote_t4_gender "{\bf Notes:} This table is based on Table 4 in Hussam et al. (2022b) and introduces interactions with a dummy variable taking the value of 1 when it is male entrepreneurs instead of female entrepreneurs who are ranked."
	
	global footnote_t2_gender "{\bf Notes:} This table is based on Table 2 in Hussam et al. (2022b) and introduces interactions with a dummy variable taking the value of 1 when it is male entrepreneurs instead of female entrepreneurs who are ranked."
	

	
